I think a more likely scenario is that GPU time in the cloud will be very cheap. If LLM use/profitability doesn't follow the expected path, new deep learning ap

I think a more likely scenario is that GPU time in the cloud will be very cheap. If LLM use/profitability doesn't follow the expected path, new deep learning applications could be heavily subsidised.